Extracting or Guessing? Improving Faithfulness of Event Temporal Relation Extraction

TL;DR

This paper enhances the faithfulness of event temporal relation extraction by mitigating training biases, incorporating tense information, and improving uncertainty estimation, leading to more accurate and reliable TempRel predictions especially under distribution shifts.

Contribution

It introduces counterfactual bias mitigation, tense-aware representations, and Dirichlet-based uncertainty estimation to improve TempRel extraction fidelity.

Findings

Better TempRel extraction accuracy on MATRES, MATRES-DS, and TDDiscourse datasets.

Improved model calibration and selectivity in relation predictions.

Enhanced robustness under distribution shifts.

Abstract

In this paper, we seek to improve the faithfulness of TempRel extraction models from two perspectives. The first perspective is to extract genuinely based on contextual description. To achieve this, we propose to conduct counterfactual analysis to attenuate the effects of two significant types of training biases: the event trigger bias and the frequent label bias. We also add tense information into event representations to explicitly place an emphasis on the contextual description. The second perspective is to provide proper uncertainty estimation and abstain from extraction when no relation is described in the text. By parameterization of Dirichlet Prior over the model-predicted categorical distribution, we improve the model estimates of the correctness likelihood and make TempRel predictions more selective. We also employ temperature scaling to recalibrate the model confidence measure after bias mitigation. Through experimental analysis on MATRES, MATRES-DS, and TDDiscourse, we demonstrate that our model extracts TempRel and timelines more faithfully compared to SOTA methods, especially under distribution shifts.